Mapping crossover frequency within the Arabidopsis NBS-LRR gene clusters

NBS-LRR genes frequently cluster in plant genomes [16]. For example, the majority of the Arabidopsis NBS-LRR domain gene family (114/197) are located in clusters on the south arms of chromosomes 1 and 5 (Fig 1 and S1 Table), which also correspond to regions of disrupted synteny between A.thaliana and A.lyrata chromosomes [16,36]. We constructed phylogenies using NBS domain amino acid sequences and compared the resulting trees to the physical location of NBS-LRR genes on chromosomes 1 and 5 (Fig 1C and 1D). This indicated that both tandem and dispersed NBS-LRR gene duplications contribute to cluster organization, consistent with previous analyses [16,36]. To map crossovers experimentally at high resolution within the clusters we designed a double-selection strategy (Fig 1A). This utilizes Syngenta Arabidopsis Insertion Library (SAIL) T-DNA insertions that confer glufosinate resistance in the Columbia (Col) accession, hereafter named BAR (Basta Resistant) lines [68], and Singapore Gene Trap (SGT) Ds insertions that confer kanamycin resistance in the Landsberg erecta (Ler) accession, hereafter named KAN (Kanamycin Resistant) lines (Fig 1A and 1B) [68,69]. Single copy KAN and BAR insertions that flank the NBS-LRR clusters on chromosomes 1 and 5 were identified using Southern blotting and genetic analysis (Figs 1 and S1). KAN-BAR insertion pairs used for crossover double-selection define the MAJOR RESISTANCE CLUSTER1 (MRC1, 3.87 Mb, SGT_5517 and SAIL_148_D06) and MAJOR RESISTANCE CLUSTER5 (MRC5, 4.69 Mb, SAIL_309_G06 and SGT_5183) map regions (Fig 1A and S2 and S3 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Meiotic recombination hotspots and coldspots within MAJOR RESISTANCE CLUSTER1 and MAJOR RESISTANCE CLUSTER5.

(A) Crossing scheme used to double-select MRC crossovers. KAN (SGT) Ler chromosomes are shown in red and BAR (SAIL) Col chromosomes are shown in green. Triangles represent insertions and are labeled with the resistance they confer (KAN = kanamycin, BAR = glufosinate). (B) Diagram illustrating MRC genotype expectations in double-resistant backcross progeny, depending on crossover location. A representative ethidium-stained genotyping gel for a Col/Ler single nucleotide polymorphism (SNP) dCAPs marker is shown beneath [70]. Colored stars above the genotyped samples match the chromosome diagrams above. (C-D) Phylogenetic trees generated using NBS domain amino acid sequences are plotted, with branch tips connected to the corresponding gene physical location. (E-F) Physical maps of chromosomes 1 and 5 with red dots indicating NBS-LRR gene positions (S1 Table). The selected MRC regions are indicated by black vertical lines and KAN and BAR symbols. (G-H) Recombination rate (cM/Mb) throughout the genotyped MRC1 and MRC5 regions (S2 and S3 Tables). Red x-axis ticks indicate NBS-LRR gene positions (S1 Table). The dotted horizontal line indicates the male Col/Ler chromosome average recombination rate [71]. The RPP8 and RPS6 coldspot locations are indicated beneath the MRC5 map. (I) Crossover frequency within the HRG1 hotspot interval (S4 Table). Arrows represent genes, with the R gene highlighted in red. (J) As for (I) but for HRG2/HRG3 (S5 Table). (K-L) Historical recombination rates across HRG1 and HRG2/HRG3. Crossover frequency (cM/Mb) estimates (LDhat) from 80 Eurasian and 180 Swedish accessions are shown in blue and red respectively [72–74]. X-axis ticks indicate SNP positions used for analysis. Horizontal dotted lines indicate chromosome mean historical recombination values. Arrows represent genes, with the NBS-LRR genes highlighted in red.

https://doi.org/10.1371/journal.pgen.1006179.g001

Paired MRC KAN-BAR insertions were crossed to generate double kanamycin-glufosinate resistant F₁ trans-hemizygotes, which were then backcrossed as males to wild type Col females (Fig 1A). Backcross progeny were grown on kanamycin-glufosinate media and the ratio of double-resistant to sensitive individuals was used to measure genetic distances, according to the formula cM = 100×(2×(N_KB/N_total)) (Fig 1A and Table 1), where N_KB is the number of kanamycin-glufosinate double-resistant progeny and N_total is the total number of progeny screened. Genotyping of double-resistant individuals with Col/Ler Simple Sequence Length Polymorphisms (SSLP) markers was used to confirm that they had experienced a crossover within the selected MRC region (Fig 1B and S2 and S3 Tables). Average recombination rates within the MRC regions were comparable to the male Col/Ler genome average 4.82 cM/Mb (MRC1 5.89 cM/Mb, MRC5 4.18 cM/Mb) (Table 1 and S2 and S3 Tables) [71]. To analyze fine-scale recombination patterns within the MRC regions a subset of the double-selected MRC1 (n = 725) and MRC5 (n = 714) recombinant individuals were chosen for further analysis by internal genotyping (Fig 1 and S2 and S3 Tables). Double-resistant crossover progeny were genotyped for 93 and 90 Col/Ler SNPs within the clusters to identify MRC1 and MRC5 marker intervals containing crossovers, with a mean inter-marker distance of 46.5 kb (Fig 1 and S2 and S3 Tables). Analysis of crossover patterns in the MRC maps revealed intervals with both significantly higher and lower crossovers than expected if random (MRC1, X² = 441.45 d.f. = 93 P = <2.2×10⁻¹⁶; MRC5, X² = 337.29 d.f. = 90 P = <2.2×10⁻¹⁶), including intervals containing R genes (Fig 1 and S2 and S3 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Double-selection of MRC1 and MRC5 crossovers.

KAN-BAR F₁ trans-hemizygotes were backcrossed to Col wild type female flowers and progeny were double-selected for kanamycin and glufosinate resistance. The numbers of double-resistant versus sensitive backcross seedlings are listed. The physical distances (megabase, Mb) were calculated between the KAN and BAR insertion sites relative to the TAIR10 Col reference assembly [75]. Genetic distance (cM) is calculated as 100×(2×(N_KB/N_total)), where N_KB is the number of Kan-Bar double-resistant backcross progeny and N_total is the total number of backcross progeny screened (Double-Resistant+Sensitive).

https://doi.org/10.1371/journal.pgen.1006179.t001

NBS-LRR gene hotspots and coldspots within MRC1 and MRC5

The MRC1 interval with the highest recombination contains a functionally uncharacterized TIR-NBS-LRR gene (At1g65850), which for the purposes of this study we designate HOT R GENE1 (HRG1) (Fig 1G and S2 Table). HRG1 shares high sequence identity with RPP1 and DM2, which confer resistance to Hylaperonospora parasitica and cause hybrid necrosis respectively [32,76]. The HRG1 map interval contained 18 crossovers, which we fine-mapped by genotyping using derived Cleaved Amplified Polymorphisms (dCAPs) markers (Fig 1I) [70]. We observed a concentration of crossovers (14 of 18) within HRG1, with a peak recombination rate of 182.87 cM/Mb (Fig 1I and S4 Table), compared to the male Col/Ler genome average of 4.82 cM/Mb [71]. Similarly, the interval in MRC5 with highest crossover frequency contains a tandem pair of TIR-NBS-LRR genes (At5g46260 and At5g46270), which we designate HOT R GENE2 (HRG2) and HOT R GENE3 (HRG3) (Fig 1H and S3 Table). HRG2 and HRG3 are also functionally uncharacterized, but show high sequence similarity with RPS6 and RAC1, which confer resistance to Pseudomonas syringae and Albugo candida respectively [60,77]. Fine-mapping crossovers using additional dCAPs markers again revealed crossovers concentrated within the HRG2 and HRG3 NBS-LRR genes (11 of 11 crossovers) (Fig 1J and S5 Table). We fine-mapped crossovers in six additional high recombination intervals in MRC1 and MRC5 that contained NBS-LRR genes (Fig 2). We observed enrichment of crossovers within NBS-LRR genes in all cases (WRR4, CW9, HRG4, HRG5, HRG6, HRG7, HRG8 and HRG9) (Fig 2 and S6–S11 Tables). Recombination rates observed at the MRC NBS-LRR hotspots were comparable to characterized plant hotspots and were detected at singleton, tandemly duplicated and complex NBS-LRR loci (Fig 2 and S4–S12 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Fine-mapping MRC NBS-LRR gene crossover hotspots.

Crossover frequency (cM/Mb) in high recombination MRC intervals containing NBS-LRR genes (S6–S11 Tables). Internal dCAPs marker positions are indicated by x-axis ticks, in addition to whether the region is located in MRC1 (Chr1) or MRC5 (Chr5). NBS-LRR genes are indicated by red arrows and non-NBS-LRR genes by black arrows. The chromosome average recombination rate for Col×Ler male backcrosses is indicated by horizontal, dotted blue lines [71]. Vertical black lines indicate the location of SNP markers genotyped in MRC maps (S2 and S3 Tables). NBS-LRR domain gene names are printed within the plot and symbols used to indicate which gene arrows they correspond to.

https://doi.org/10.1371/journal.pgen.1006179.g002

To test for MRC intervals with significantly different crossover rates, we constructed 2×2 contingency tables and performed X² tests using the observed crossover and non-recombinant counts, and those expected at random based on the physical size of each interval. Finally, we applied a false discovery correction, due to multiple testing [78]. Out of 185 intervals in MRC1 and MRC5, 22 were significantly different from the null expectation (S2 and S3 Tables). Ten intervals showed higher recombination, which included HRG1, HRG2, HRG3, HRG4 and HRG5 (S2 and S3 Tables). Twelve intervals showed lower recombination, which included intervals containing RRS1B-RPS4B, the RPS6 cluster and a cluster of 4 uncharacterized CC-NBS-LRR genes (At1g58807, At1g58848, At1g59124 and At1g59218) (S2 and S3 Tables). However, after correction for multiple testing, only HRG1 and the CC-NBS-LRR cluster remained significantly different (S2 and S3 Tables) [78]. To test whether intervals containing NBS-LRR genes showed different recombination rates compared to those that did not, we compared observed crossover counts to those expected at random in proportion to physical distance, using a X² test. MRC intervals that contained NBS-LRR genes did not have significantly different crossover counts compared to those that lacked NBS-LRR genes (P = 0.418).

As a further test of recombination in NBS-LRR genes compared to other genes, we analyzed 1,230 crossovers mapped throughout the genome by genotyping-by-sequencing of 192 Col×Ler F₂ individuals (S13 Table) [79,80]. The crossovers were mapped to a mean resolution of 947 bp, representing 1,164,432 bp in total, which significantly overlapped with genes (Block-bootstrap Z = 4.54, P = 2.77×10⁻⁶). To compare gene groups we sampled from the non-NBS-LRR genes using parameters that approximated the clustering structure of the NBS-LRR genes (S2A and S2B Fig). We repeatedly sampled and tested for overlap between samplings and crossovers, requiring at least one base pair in common to score an overlap. Values were also length-normalized by rescaling the observed data by the total length in base pairs of the sample or gene set (S3C–S3F Fig). For both NBS-LRR and non-NBS-LRR genes the observed proportion of genes overlapping a crossover lie within those observed for the sampled data, usually near the distribution’s center (S3C–S3F Fig). Therefore we conclude that the number of NBS-LRR and non-NBS-LRR genes overlapping crossovers is not substantially different, and that both groups show the presence of hotspots and coldspots.

If recombination hotspots within NBS-LRR genes have persisted in the species, then decay of linkage disequilibrium (LD) between natural genetic polymorphisms is expected [81,82]. For example, the presence of all four possible gametes for pairs of linked polymorphisms (i.e. AB, aB, Ab and ab) can be used as a measure of historical recombination [81,82]. We therefore analyzed single nucleotide polymorphisms (SNPs) from 80 Eurasian and 180 Swedish Arabidopsis accessions using LDhat [72–74]. Consistent with our experimental crossover measurements HRG1, HRG2 and HRG3 show elevated historical recombination (Fig 1K and 1L) [72–74]. We also observed intervals containing R genes, like RPP8, with low experimental recombination, but which exhibited high historical recombination frequency (Fig 1H and S1–S3 Tables). The suppressive effect of structural polymorphism on recombination is likely to contribute to these patterns. For example, RPP8 is tandemly duplicated in Ler relative to Col [50], and the RPS6 cluster shows complex Col/Ler structural polymorphism [77]. In summary, although overall MRC recombination rates were close to the genome average, we observed heterogeneity in local crossover frequency and the presence of NBS-LRR gene associated hotspots and coldspots.

Intragenic recombination hotspots in the RAC1 resistance gene

To investigate crossover variation within a single R gene we used pollen-typing [37,83,84], which is analogous to mammalian sperm-typing approaches [85,86]. This technique uses two rounds of nested, allele-specific PCR to specifically amplify and titrate either crossover or parental (non-recombinant) molecules from F₁ (e.g. Col/Ler) pollen DNA (Fig 3A and S3 Fig). Importantly, crossover products are not amplified from Col/Ler F₁ somatic leaf DNA under the same PCR conditions (Fig 3A). This demonstrates that template switching is not generating crossover molecules during PCR amplification. The concentrations of parental and crossover molecules, measured by titration, are used to calculate recombination rate within the amplified region [37,83–85]. Amplification from single molecules containing crossovers, followed by Sanger sequencing, is used to identify internal recombination locations, to the resolution of individual polymorphisms (Fig 3A) [37,83,84]. Between 100–200 crossover molecules are Sanger sequenced to map the topology of recombination within an amplified region and calculate crossover frequency (cM/Mb) for each marker interval (Fig 3A) [37,80,83,84]. Physical distances analyzed by pollen-typing are typically between 5–10 kb [37,80,83,84].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Intragenic crossover hotspots within the RAC1 resistance gene.

(A) Diagram illustrating pollen-typing methods used to fine-map plant crossover hotspots. Pollen is collected from Col/Ler F₁ plants and genomic DNA is used for two rounds of nested allele-specific PCR amplifications with either crossover (ASO-Ler+ASO-Col) or parental (ASO-Col+ASO-Col) allele-specific oligonucleotide (ASO) primer combinations. An ethidium-bromide stained agarose gel is shown with crossover-specific amplifications from Col/Ler F₁ pollen, which are not observed when using Col/Ler F₁ leaf (somatic) DNA as a template. Crossover positions are identified using either Sanger sequencing from single crossover molecule amplifications, or paired-end sequencing of crossover molecule libraries. (B) Historical crossover frequency over RAC1 estimated by LDhat analysis of polymorphisms from 80 Eurasian (blue) and 180 Swedish (red) accessions [72,73]. X-axis ticks indicate SNP positions used for analysis. Horizontal dotted lines indicate chromosome mean values. Black arrows represent genes and green horizontal lines indicate transposable elements. Black vertical lines indicate RAC1 TSS and TTS positions. (C) Crossover frequency within the RAC1 amplicon measured by crossover and parental molecule titration, followed by single crossover molecule Sanger sequencing. The positions of RAC1 and the adjacent gene (At1g31550) are indicated by horizontal black arrows above the plot. Exons are indicated by horizontal black lines, with sequences encoding RAC1 TIR (green), NB-ARC (red) and LRR (blue) domains highlighted as coloured lines. For the NB-ARC domain, the position of the Walker A, Walker B, ARC1 and ARC2 motifs are shown. Red x-axis ticks show Col/Ler SNP positions. Blue vertical lines indicate the positions of gene TSS and TTS from TAIR10 representative gene models.

https://doi.org/10.1371/journal.pgen.1006179.g003

To select candidate hotspot R genes for pollen-typing analysis we analyzed historical recombination rates [37,72–74]. We used LDhat to analyse SNPs from 80 Eurasian and 180 Swedish Arabidopsis accessions [72,73], and observed that RPP8 and RAC1 are the R genes with highest LD-based recombination estimates (S1 Table). However, the RPP8 locus shows structural polymorphism between accessions, including between Col and Ler, making it unsuitable for pollen-typing analysis [50]. In contrast, RAC1 is present as a TIR-NBS-LRR singleton in both Col and Ler and so was selected for further study. RAC1 alleles confer resistance to Hyaloperonospora arabidopsidis and Albugo candida and shows high sequence identity with HRG2, HRG3 and RPS6 [60–62]. RAC1 resistance has been mapped using the Acem1 Albugo isolate, which both Col and Ler are susceptible to [60–62]. We amplified a 9.45 kb region, spanning RAC1 and an adjacent gene (At1g31550) that encodes a putative lipase, and measured a recombination rate of 16.99 cM/Mb, which is comparable to known plant crossover hotspots (Fig 3C and S12 and S14 Tables). After sequencing 181 single molecules containing crossovers we observed highest recombination within the RAC1 open reading frame, with a peak intragenic crossover frequency of 117 cM/Mb (Fig 3C and S14 and S15 Tables). RAC1 recombination was highest in the 5' exons that encode the TIR and NBS domains, and in a 3'-region encoding LRR domains (Fig 3C). This pattern is consistent with high intragenic recombination rates observed at the 5' and 3' ends of plant genes [37,63,64,87]. In this regard it is important to note that residues critical for resistance function have been mapped to all three domains (TIR, NBS and LRR) in other R genes [58,88–91]. The open reading frame of the adjacent gene At1g31550 also showed substantial crossovers (Fig 3C). Together these data demonstrate that RAC1 contains strong historical and contemporary crossover hotspots.

To analyze greater number of recombination events, we mass-amplified approximately 1,000 RAC1 crossover molecules using 672 independent PCR reactions (Fig 3A). Each reaction was estimated to contain between 1 and 2 crossover templates, based on our previous titration experiments (S14 Table). The resulting crossover amplification products were then pooled, sonicated and used to generate sequencing libraries (Fig 3A). Paired-end sequencing was performed and read pairs were separated and aligned to the Col and Ler parental RAC1 templates, allowing only exact matches (S16 Table). Reads were filtered for those that aligned to one parent only, and for pairs with a centromere-proximal read mapping to Col and a centromere-distal read mapping to Ler, on opposite strands (S16 Table). Read pairs matching these criteria are consistent with the crossover allele-specific oligonucleotide (ASO) configuration used during PCR amplification (Fig 3A). These filters yielded 182,909 total crossover read pairs (S16 Table). We assigned recombination values to marker intervals between crossover read pairs, weighted by the number of intervals and their widths, and normalized by the total number of mapping reads (Fig 4A). This revealed a similar recombination pattern to that observed from single molecule Sanger sequencing, with high recombination overlapping the RAC1 TIR, NB-ARC and LRR domain exons (Figs 3C and 4A). Crossover read pair analysis also confirmed substantial recombination activity within the adjacent gene At1g31550 (Fig 4A). Together these data demonstrate focused intragenic recombination in RAC1 overlapping exon sequences [60–62].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RAC1 crossovers measured by high-throughput sequencing, nucleosome occupancy and CTT-repeat motifs.

(A) Normalized crossover read pair values are plotted relative to the TAIR10 reference sequence, with the position of Col/Ler polymorphisms indicated by red x-axis ticks. Labelling is as for Fig 3C. (B) The sequence of the RAC1 CTT-repeat motif is printed and connected to its physical location in (A) with dashed lines. Note the sequence is the reverse complement relative to the orientation of plotting in (A). The presence of CTT-motifs in RAC1-related paralogs in the Col-0 accession are shown in a multiple sequence alignment, where gene start codons (ATG) are in bold type. The asterisks beneath the alignment indicate identical positions. (C) Normalized nucleosome occupancy values (MNase-seq) plotted as for (A).

https://doi.org/10.1371/journal.pgen.1006179.g004

We investigated the extent to which chromatin and DNA sequences associate with crossover frequency at the RAC1 hotspot. Previous studies have identified degenerate CTT and CCN repeat motifs associated with Arabidopsis hotspots, enriched at the 5′-end of genes [37,63,64]. We observed a CTT-motif (CTTCGTCATCTTCTTCT) almost immediately downstream of the RAC1 start codon, within the TIR-domain encoding exon that shows high recombination rate (Fig 4A and 4B). CTT-motifs are also observed in RAC1-related NBS-LRR genes (Fig 4B). To investigate association of CTT-motifs with recombination we divided NBS-LRR genes located within the MRC maps into those in intervals with higher or lower crossover frequency, compared to the male Col×Ler genome average (4.82 cM/Mb) [71]. 62% (21 of 34) of high recombination NBS-LRR genes showed at least one match to a 16-mer CTT-repeat motif that we previously identified as enriched at historical crossover hotspots (S17 Table) [37]. In contrast, significantly fewer (33%, 16 of 49) low recombination NBS-LRR genes showed matches to this CTT-motif (S17 Table) (X² P = 0.016). This analysis further demonstrates association between high crossover recombination and CTT-repeat motifs in Arabidopsis [37,39,63,64]. To analyse chromatin structure we performed micrococcal nuclease digestion of wild type (Col-0) chromatin and gel purified the resulting ~150 bp mononucleosomal DNA band [92]. This DNA was used to generate a library and paired-end sequencing performed, followed by analysis using the nucleR package [93]. We observed that intergenic regions within the pollen typing amplicon, and RAC1 intron 4, showed relatively low nucleosome occupancy (Fig 4C) [92]. Highest crossover rates correlated with nucleosome-occupied exon sequences within RAC1 and At1g31550 (Fig 4A and 4C). Together these analyses demonstrate associations between RAC1 crossover frequency, DNA sequence motifs and chromatin structure.

R gene structural variation and crossover modification by genetic background

The HRG1, HRG2-HRG3 and RAC1 NBS-LRR hotspots show historical and experimental recombination rates above the genome average (Figs 1, 3, S4, S5, S14 and S15 Tables), whereas other R genes differed between these two measures. For example, RPP8 shows high historical recombination, yet low experimental crossover frequency (Fig 1H and S3 Table). RPP8 is a singleton R gene in the Col reference genome, but in Ler occurs as a tandemly duplicated pair with RPH8A [50]. Sequence analysis indicates that the Col RPP8 haplotype is derived from a gene fusion event between RPP8 and RPH8A [50]. We performed Southern blotting and hybridization analysis and confirmed the RPP8/RPH8A Col/Ler structural polymorphism (Fig 5A) [50]. The observed band sizes are also consistent with a recent Ler genome assembly generated by PacificBiosystems (Fig 5B) [94]. Hence, low RPP8 recombination in the Col/Ler MRC5 map may be explained due to structural heterozygosity inhibiting crossovers [95,96]. Similarly, the RPS6 cluster shows structural polymorphism between Col and Ler and an absence of crossovers in MRC5, despite showing high sequence similarity to the hotspot R genes RAC1, HRG2 and HRG3 (Figs 1H, 5C and 5D and S3 Table) [77]. In contrast the hotspot R genes HRG1, HRG2, HRG3 and RAC1 show low levels of structural polymorphism between accessions, including between Col and Ler (Figs 5C and 5D and S4). For the R genes RPP1 and RPP4 Southern blotting and hybridization analysis revealed complex patterns of structural polymorphism that did not obviously correspond to the PacBio Ler assembly (S5 and S6 Figs) [94]. Hence, for these R genes structural polymorphism is also likely to be inhibitory to recombination, when divergent haplotypes are heterozygous.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. NBS-LRR gene structural diversity between Arabidopsis accessions.

(A) Southern blotting and hybridization analysis of RPP8 in the indicated accessions. Cross-hybridizing NBS-LRR genes in Col-0 predicted from the TAIR10 reference assembly are labelled in black. In Col-0 RPP8 is detected as a 4.3 kb EcoRV band, and two 5.5 and 6.1 kb bands in Ler-0. Bands corresponding to Ler-0 RPP8 and RPPH8A genes are labelled in blue. (B) Diagram showing annotated structural variation at RPP8 between Col-0 (TAIR10) and Ler-0 (PacBio) assemblies [94]. Regions predicted to cross-hybridize with the probe are indicated in red, in addition to the position of EcoRV sites. (C) and (D) are as for (A) and (B), but probing for RAC1. The RAC1 probe cross-hybridizes with a linked cluster of NBS-LRR genes, including RPS6. In Col-0 the RPS6 cluster includes three NBS-LRR genes (At5g46490, At5g46510, At5g46520) that are present as a single gene (At5g46490) in Ler-0; such that in Col-0 three bands of 3.7, 4.5 and 4.7 kb are detected after EcoRV and HindIII digestion, and one 2.8 kb band in Ler-0. The Ler-0 band is labelled in blue.

https://doi.org/10.1371/journal.pgen.1006179.g005

To investigate the extent of inter-cross variation in NBS-LRR cluster recombination we used segregation of linked Fluorescent Tagged Line (FTL) T-DNA insertions [67,97]. Inheritance of FTLs in pollen can be monitored via fluorescence microscopy or flow cytometry (Fig 6A) [67,97]. Pollen grains collected from FTL/++ hemizygotes consists of four fluorescent classes that represent parental (no color and red+yellow) versus crossover recombinant (red or yellow alone) chromosomes (Fig 6B). The ratios between these fluorescent pollen classes are used to calculate genetic distance between the FTL T-DNA insertions (Fig 6A and 6B) [66,67,97]. Pollen fluorescence was analyzed using a flow cytometer allowing >10,000s of pollen grains to be scored per individual plant (Fig 6C and S18 Table) [67]. We used FTL lines that define a 4.92 Mb interval called I5a, which has a 2.09 Mb overlap with the MRC5 region and contains 33 out of 70 chromosome 5 NBS-LRR genes (Fig 6A and S1 Table). We crossed I5a (generated in the Col-0 background) to 18 Arabidopsis accessions representing global diversity and measured genetic distance in replicate F₁ plants, scoring ~30,000 pollen grains per individual (Fig 6 and S18 Table). We observed 3 F₁ crosses with significantly higher I5a recombination and 15 F₁ crosses with significantly lower I5a recombination, relative to Col/Col homozygotes (general linearized model P<2.0×10⁻¹⁶) (Fig 6D and S18 Table). This demonstrates the sensitivity of NBS-LRR cluster recombination rates to genetic background, which is consistent with observed cis and trans modification of crossover frequency by natural genetic variation in plants [97–105]. Therefore, Arabidopsis genetic variation has a significant modifying effect on crossover frequency within the NBS-LRR clusters.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. NBS-LRR cluster recombination rate is modified by natural genetic variation.

(A) Physical map of chromosome 5 with NBS-LRR gene positions indicated by red dots. The positions of FTL T-DNAs that define I5a are indicated by vertical blue lines, with triangles indicating the colour (red or yellow) of fluorescent protein encoded. Beneath is a diagram illustrating the generation of FTL hemizygous F₁ plants and scoring of crossover recombination in pollen. Crossover within I5a leads to pollen with red or yellow fluorescence alone, whereas the parental classes are either without fluorescence or show both red and yellow. (B) Fluorescence micrograph of pollen from I5a/++ hemizygotes, showing the presence of four fluorescent classes. (C) A representative flow cytometry plot of pollen from I5a/++ hemizygotes. Points represent individual pollen grains measured for yellow (FL1 YFP) and red (FL2 dsRed2) fluorescence. (D) I5a genetic distance (cM) measured from individual F₁ plants (S18 Table). Mean values are shown in red. The accession crossed to is indicated on the x-axis. The Col/Col homozygous sample is labelled and indicated with an arrow.

https://doi.org/10.1371/journal.pgen.1006179.g006

NBS-LRR gene identification and analysis

We combined published annotations of the Arabidopsis NBS-LRR family with our own manual searches and selected a set of 197 genes for analysis (S1 Table) [16,35]. To analyse phylogenetic relationships NBS domain amino acid sequences were aligned using ClustalOmega and gaps treated as missing data. The resulting alignment was analysed using the PhyML 3.0 webserver (http://www.atgc-montpellier.fr/phyml/) with 500 bootstraps performed for tree construction [135]. To search for CTT-repeat motifs associated with NBS-LRR genes, a position weight matrix from a historical hotspot-enriched motif was matched at 80% to 1 kb windows surrounding NBS-LRR gene transcriptional start sites (TSSs).

Double-selection and analysis of MRC1 and MRC5 crossovers

To select for crossovers with the NBS-LRR gene clusters we followed an identical strategy to that used previously for the TEL1 interval (Fig 1A) [37]. SAIL T-DNA insertions conferring glufosinate resistance in the Col accession (hereafter named BAR), and Ds insertions conferring kanamycin resistance in the Ler accession (hereafter named KAN), were chosen with insertion sites flanking the NBS-LRR gene clusters on the south arms of chromosomes 1 and 5 [68,69]. Southern blotting and hybridization was used to identify single copy KAN and BAR insertion lines (S1 Fig). Single copy lines were tested that (i) self-fertilization of homozygotes gave ~100% resistant progeny, (ii) self-fertilization of hemizygotes gave ~75% resistant progeny and (iii) backcrossing hemizygotes to wild type gave ~50% resistant progeny (S1 Fig). After screening we selected SGT_5517 (KAN) and SAIL_148_D06 (BAR) for chromosome 1 and SAIL_309_G06 (BAR) At5g49220 and SGT_5183 (KAN) for chromosome 5. KAN and BAR lines were then crossed to generate trans-hemizygote F₁ individuals, which were backcrossed as males to Col-0 (Table 1). Backcross progeny were germinated on media containing kanamycin and glufosinate to select for individuals that experienced a crossover between the KAN and BAR insertions. Genetic distance was calculated as: cM = 100×(2×(N_KG/N_total)) (Table 1), where N_KG is the number of Kan-Bar double-resistant backcross progeny and N_total is the total number of backcross progeny screened. We used Col/Ler SSLP genotyping markers to confirm that surviving progeny had experienced a crossover between the MRC insertions. Recombinants were selected using SSLP markers to restrict analysis to the MRC1 and MRC5 intervals (S2 and S3 Tables). Genomic DNA from these individuals was then genotyped using sets of Col/Ler SNP markers within MRC1 and MRC5 (KBiosciences). To test whether crossover distributions were significantly different from random we performed a 2×n X² test, where n = the number of map intervals. To test the significance of individual intervals X² tests were performed between the observed crossover and non-recombinant counts with those expected at random using 2×2 contingency tables (S2 and S3 Tables). Random expectations were obtained by proportionally allocating crossovers to intervals according to their physical length, i.e. an interval twice as long would be expected to receive twice as many crossovers. Finally, a false discovery correction was applied due to multiple testing [78].

Genotyping by sequencing and crossover analysis

DNA was extracted using CTAB and used to generate sequencing libraries as described [79,80], with the following modifications. DNA was extracted from 3 rosette leaves of 5 week old plants and 150 ng of DNA used as input for each library. DNA shearing was carried out for 20 minutes at 37°C with 0.4U of DNA Shearase (Zymo research). Each set of 96 libraries was sequenced on one lane of an Illumina NextSeq500 instrument (300-cycle Mid Output run). Sequencing data was analysed to identify crossovers as previously reported, using the TIGER pipeline [79,80]. The fastq files for these data have been uploaded to ArrayExpress accession E-MTAB-4657.

To compare gene groups we sampled from the non-NBS-LRR genes by taking n blocks of length b randomly distributed in the genome and sampled a proportion p of genes within each block. We compared the distributions of the distance between the start coordinates of the sampled genes and the physical widths of the sampled genes, to those observed in the NBS-LRR genes, performing parameter sweeps on n, b and p, in order to minimize the sum of the Kolmogorov-Smirnov statistics observed for each distribution. The minimal difference between the distributions was observed for 220 blocks of 35,000 base pairs and sampling 20% of the genes in each block. Using these parameters, we obtained an approximation to the clustering structure of the NBS-LRR genes (S2A and S2B Fig). We then repeatedly sampled (3,200 times) using these parameters and tested for overlap between samplings and crossovers, requiring at least one base in common to score an overlap. Values were length-normalized by rescaling the observed data by the total length in base pairs of the sample or gene set (S3C–S3F Fig). The observed overlap between NBS-LRR and non-NBS-LRR genes with crossovers were then compared to the sampled data (S3C–S3F Fig).

RAC1 hotspot pollen-typing and crossover sequencing

The RAC1/RPP9 gene and surrounding region were PCR amplified and Sanger sequenced from the Ler-0 line used in our laboratory to identify all associated SNPs and indel polymorphisms. Using this information we designed allele specific oligonucleotides (ASOs) that specifically amplify from either Col or Ler genomic DNA templates, as described previously [37,83,84,136] (S1 Fig and S19 Table). For RAC1 two nested ASO amplifications were performed with the following sets of ASO primer pairs to amplify either crossover (Ler-Col) or parental (Col-Col) molecules:

1. 1^st Crossover amplification = KC459 + KC418
2. 1^st Parental amplification = KC493 + KC418
3. 2^nd Crossover amplification = KC465 + KC417
4. 2^nd Parental amplification = KC495 + KC417

The same PCR conditions were used for 1^st and 2^nd PCRs, which were:

94°C for 2’30”,

then 5 cycles of {94°C for 30”, 65°C (decrease by 0.4°C per cycle) for 45”, 68°C for 8’00”},

then 25 cycles of {94°C for 20”, 63°C for 30”, 68°C for 8’00”},

then {68°C for 10’00”}.

Recombination rate was estimated by titration of crossover and parental molecule concentrations from F₁ pollen genomic DNA [37,83,84,136]. DNA concentrations were identified which gave approximately half crossover positive amplifications, where by Poisson approximation the majority are from single molecules [37,83,84,136]. Single crossover molecules were amplified and treated with exonuclease I (NEB, M0293) and shrimp alkaline phosphatase (Amersham, E70092), and then directly Sanger sequenced to identify internal crossover locations.

Based on crossover molecule concentrations measured by titration, we performed mass amplification (672 reactions) of crossovers, such that most contained between 1–2 crossover templates. The resulting amplification products were pooled, concentrated by isopropanol precipitation and gel purified. ~2 μg of DNA in 100 μl of TE (10 mM Tris, 1mM EDTA, pH 8.0) buffer was then sonicated to a size range of 300–400 bp using a Bioruptor (Diagenode). Sonicated DNA was gel purified, end-repaired and used to generate libraries (Tru-seq, Illumina). These libraries were sequenced on an Illumina nextSeq instrument using paired-end 75 bp reads. Paired reads were separated and aligned to the Col or Ler parental sequences using bowtie allowing only exact matches [137]. Reads were filtered for those that uniquely aligned to one parental sequence only. We then filtered for read pairs that had a centromere-proximal match to Col and a centromere-distal match to Ler on opposite strands, in order to identify crossover read pairs (S16 Table). These criteria are consistent with the ASO configuration used during allele-specific PCR amplification. A value of 1 was divided by the distance (bp) between each crossover read pair. For example, if a given pair of crossover reads are 200 bp apart, then the fractional value per bp is 1/200 = 0.005. Further consider 2 SNPs located between these crossover reads that divide this region into 3 intervals of 10, 50 and 140 bp. Each interval would then be assigned values of 10*0.005, 50*0.005 and 140*0.005 respectively. This process is repeated for all read pairs and final values normalized by the total number of crossover read pairs. The RAC1 pollen-typing fastq files have been uploaded to ArrayExpression accession E-MTAB-4556.

Analysis of historical crossover frequency

Diallelic SNPs with less than 10% missing data were selected from 80 Eurasian and 180 Swedish accession 1,001 Genomes Project datasets [72,73]. SNPs overlapping repeats and the centromeres were masked due to difficulties in assembling these regions, as performed previously [37]. SNP data was analysed using LDhat run parameters described previously, but using a θ = 0.001 look-up table [37,74]. LDhat recombination maps were filtered by setting regions where 4N_er >100, or the distance between SNPs >50 kb, to zero recombination. Population-scaled recombination rates were rescaled to cM/Mb by regression onto a consensus experimental F₂ genetic map, as described previously [37].

Micrococcal nuclease sequencing

Chromatin from 1 gram of wild type (Col-0) flowers (unopened buds, prior to floral stage 12) was isolated and digested with 0.05 units of mirococcal nuclease (MNase, NEB) in reaction butter (10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 1 mM EDTA, 4 mM CaCl₂) at 37°C for 15 minutes, with occasional vortexing. The digested nucleosomal DNA was then treated with proteinase K (20 μg/μl) and phenol/chroloform extraction performed, followed by ethanol precipitation. The digested DNA was separated using gel electrophoresis through an ethidium-stained agarose-TBE gel and the ~150 bp nucleosomal band was gel purified. This DNA was end-repaired and used to generate Illumina sequencing libraries (Tru-seq) and subjected to paired-end 100 bp sequencing on a Hi-seq instrument. Reads were mapped using Bowtie2 to the TAIR10 assembly and normalized by library size. Nucleosome occupancy was analysed using the nucleR package [93]. The MNase-seq fastq files have been uploaded to ArrayExpression accession E-MTAB-4556.

NBS-LRR gene Southern blotting and hybridization

1.5 μg of genomic DNA was digested using 100 units of restriction enzyme (NEB) in a final volume of 200 μl for 15 hours at 37°C. Digested DNA samples were isopropanol precipitated, resuspended in TE and electrophoresed in 0.8% agarose-TBE gels. Gels were blotted onto a positively charged nylon membrane (Hybond-N+, Amersham Biosciences), which was hybridized in 0.5 M phosphate buffer, 7% w/v SDS, 1 mM EDTA (pH 8) and 1% BSA at 65°C. DNA probes to the RAC1 gene (a PCR fragment amplified with RAC1-F 5′-TGATAGAGATTGAGGCGGTCA-3′ and RAC1-R 5′-GGCTTCGTCATCTTCTTCTCAC-3′), RPP8 (a PCR fragment amplified with RPP8-F 5′-TCTTCATTCCTTCAATCTTCAGT-3′ and RPP8-R 5′-GAGATACTTGAGTTTATACGAGGCTAA-3′), HRG1 (a PCR fragment amplified with HRG1-F 5′-GGTTTGGTCATGGAAGTCGG-3′ and HRG1-R 5′-TCTTACCCGGTTTCCTTCCC-3′), RPP1 (a PCR fragment amplified with RPP1-F 5′-AGTCAGCAAGAGGAGACCAC-3′ and RPP1-R 5′-TCCAACTTAGTGCAATCATGGTC-3′) or RPP4 (a PCR fragment amplified with RPP4-F 5′-AGAAGGCAAACGCTTCACTG-3′ and RPP4-R 5′-AGCTGCCATCTCAAGGTCTT-3′) were labelled with [α-32P]-dCTP using a random priming labelling kit (Rediprime II DNA labelling system, Amersham). Blots were washed with 0.5% SSC, 0.1% SDS solution at 65°C and imaged with a PhosphoImager (Typhoon 8610, Molecular Dynamics). We compared Southern blot data with the TAIR10 Col assembly and a Ler assembly generated by Pacific Biosystems (http://blog.pacificbiosciences.com/2013/08/new-data-release-arabidopsis-assembly.html) [94].

Measurement of crossovers using flow cytometry of fluorescent pollen

Analysis of fluorescent pollen and segregation of I5a FTL T-DNAs was performed as described previously [67]. Genetic distance is calculated as cM = 100×(R5/(R3+R5)), where R5 is the number of yellow-alone pollen and R3 is the number of both yellow and red fluorescent pollen. To test whether recombinant and non-recombinant counts were significantly different between replicate groups we used a generalized linear model (GLM). We assumed the count data is binomially distributed: where Y_i represents the recombinant counts, n_i are the total counts, and we wish to model the proportions Y_i/n_i. Then: and,

Thus, our variance function is: and our link function must map from (0,1) -> (-∞, ∞). We used a logistic link function which is: where ε_i ~ N(0,σ²). Both replicates and genotypes are treated as independent variables (X) in our model. We considered P values less than 0.05 as significant.